US Quant Trading Strategies: 10 Backtested Systems Compared

Most quant write-ups show one strategy and one equity curve. This page compares ten US quant trading strategies side by side, on the same data, over the same May 2016 to October 2026 window, using one shared execution model and one shared QQQ trend gate, so what changes between them is the extra condition each adds and what each holds. Every number below comes from a real finlab backtest, and each strategy has a reproducible script.

The short version: all ten compounded faster than buying and holding the Nasdaq-100 (QQQ, 20.7% a year with dividends), from 24.3% for the conservative ETF rotation (QLD/SSO) to 51.9% for short-term reversal, and all ten had a smaller maximum drawdown than QQQ's -35.1%. All ten also beat QQQ's monthly Sharpe when both are measured the same way (0.98 for QQQ after finlab's 2% risk-free deduction), but the margins differ widely: earnings surprise leads at 1.47, while breakout/trend (1.04) and the aggressive ETF rotation (1.12) are only a little ahead. Eight of the ten also went through three robustness checks this round; seven passed all three, and breakout/trend failed one because it did not beat QQQ in the first half of the window. They are not ten unrelated systems. All ten hold leveraged growth ETFs only while QQQ is above its 200-day average with a positive 126-day return, and step aside otherwise; six of them add a breakout or factor-breadth condition on top of that gate. That is why their drawdowns cluster and why this comparison is more useful than ten separate "how I made X%" posts.

The strategy family at a glance

Each row is a distinct signal expressed through the same leveraged-ETF engine. "Signal type" separates the two halves of the family explained below; CAGR, Sharpe, and drawdown cover the full window (2016-05-02 to 2026-10-26); the out-of-sample column is the 2022-present Sharpe.

Both Sharpe columns are monthly Sharpe ratios on two conventions. The full-window column is finlab's report figure, which deducts a 2% annual risk-free rate; the strategy studies' own tables quote the figure without that deduction (momentum is 1.46 there and 1.36 here) and give the 2% figure only as a side note. The 2022- column has no risk-free deduction, as in the studies' out-of-sample tables. QQQ and SPY are computed the same way as the column they sit in, so compare down a column, not across. Benchmarks use total-return (dividend-adjusted) prices over the same dates. The leveraged ETF series already embed each fund's daily-reset behavior and expense ratio, so the volatility drag of 2x and 3x products is in the results, not assumed away. The charts keep their original labels: "09 Sector rotation" and "10 Regime timing" are the conservative and aggressive ETF rotation.

One execution model, one shared QQQ gate

The reason this is a family and not ten coincidences is the execution layer, which is the same across every strategy. Each one answers a single question every month: is the regime supportive enough to hold leveraged growth exposure? When the answer is yes, it buys one or two leveraged ETFs (TQQQ or TECL at 3x, QLD or SSO at 2x). When the answer is no, it rotates to a single defensive ETF: treasuries, gold, or a cash-like fund. The ETF-timing strategies switch cleanly between those two states. In the five stock-factor backtests the two sleeves overlap instead: in most risk-on months the book holds the defensive ETF at about 50% and TQQQ and TECL at about 25% each, so roughly half the portfolio sits in 3x funds rather than all of it (the momentum and earnings surprise studies measure this in their trade lists). Rebalancing is monthly, with a touched stop-loss on each holding: 8% for seven of the strategies, 6% for the two that hold QLD/SSO (low volatility and the conservative ETF rotation), and 10% for the aggressive ETF rotation.

The question is mostly answered by the same test. Every strategy's risk-on condition includes one QQQ gate: QQQ above its 200-day average and a positive 126-day return. Four ETF-timing strategies (02, 06, 09, 10) use only that gate and differ in what they hold and how tight the stop is: short-term reversal holds whichever of TQQQ and TECL fell more over three days, low volatility holds one 2x fund, and the two ETF rotation versions hold QLD plus SSO or the stronger of TQQQ and TECL. Breakout (07) adds one condition, a fresh 21-day QQQ high within the last five trading days. The five stock-factor strategies add a breadth threshold on top of the gate, but it is not clear how often it binds: in the momentum study the 8% breadth bar never switched the strategy off in this window, so the QQQ gate did all the switching. In a strong market the gate is open for all ten, and they often hold the same leveraged ETFs.

This design choice is visible in the data. The five stock-factor strategies share an identical -21.1% maximum drawdown, because they share the same stop, the same instruments, and the same half-defensive blend: the drawdown is a property of the engine, not the factor. It is also why publishing ten near-identical posts would be misleading. The useful unit of analysis is one engine and a menu of signals.

from finlab import data
from finlab.backtest import sim
 
# Single stocks drive the breadth signals; the ETFs are what you hold.
data.set_market("us")
stock_close = data.get("us_price:adj_close")   # adjusted: used for returns and factors
raw_close = data.get("us_price:close")         # unadjusted: used for dollar volume and the $10 floor
volume = data.get("us_price:volume")           # not split-adjusted
 
# Liquid universe for the breadth strategies: top 500 names by 60-day dollar volume.
dollar_volume = (raw_close * volume).rolling(60, min_periods=20).mean()
universe = dollar_volume.is_largest(500) & (raw_close > 10)
 
data.set_market("us_fund")
etfs = data.get("us_fund_price:adj_close")[["QQQ", "TQQQ", "TECL", "QLD", "SSO", "IEF", "GLD", "SHY"]]
 
 
# Each strategy supplies its own boolean risk_on series from a different factor.
# This clean on/off switch is what the ETF-timing strategies trade; the five
# stock-factor backtests hold a defensive/leveraged blend in risk-on months instead.
def execute(risk_on, risk_assets, n_hold=1, defensive=["IEF", "GLD", "SHY"]):
    # Risk-on: the strongest leveraged ETFs by 3-month-minus-1-month momentum.
    risk_score = etfs[risk_assets].pct_change(63) - etfs[risk_assets].pct_change(21)
    risky = (risk_score.rank(axis=1, ascending=False) <= n_hold).mul(risk_on, axis=0).astype(float)
    risky = risky.div(risky.sum(axis=1), axis=0).fillna(0)
 
    # Risk-off: the single strongest defensive ETF on the same score.
    safe_score = etfs[defensive].pct_change(63) - etfs[defensive].pct_change(21)
    safe = (safe_score.rank(axis=1, ascending=False) <= 1).mul(~risk_on, axis=0).astype(float)
    return risky.join(safe)
 
 
# Example: the QQQ-trend regime gate (strategy 10). The other ETF-timing strategies
# swap in a different risk_on definition, risk assets, or stop.
trend = etfs["QQQ"] > etfs["QQQ"].rolling(200, min_periods=100).mean()
momentum_6m = etfs["QQQ"] / etfs["QQQ"].shift(126) - 1
risk_on = trend & (momentum_6m > 0)
 
position = execute(risk_on, ["TQQQ", "TECL"], n_hold=1)
report = sim(position.loc["2016-01-01":], resample="M", stop_loss=0.10,
             touched_exit=True, trade_at_price="close")

Two kinds of signal: stock-factor breadth vs ETF timing

The family splits cleanly in two, and the split matters for how much validation each strategy carries.

Stock-factor breadth (strategies 01, 03, 04, 05, 08). These screen the full liquid US universe, the top 500 names by 60-day dollar volume, and measure how broad a factor is: what share of stocks are momentum leaders, or show positive earnings surprises, or combine cheap valuations with profitability. Broad participation is read as durable risk appetite, and the breadth threshold is added on top of the shared QQQ gate. Because there is a real stock universe underneath, each of these can be validated at the stock level first (decile sorts, rank information coefficients, a long-short spread) before any leverage is applied. The momentum write-up walks through that validation in full.

ETF timing (strategies 02, 06, 07, 09, 10). These act on the ETFs' own behavior. The risk-on decision is the QQQ gate itself (plus a fresh breakout for 07); the rest of each rule decides what to hold, such as the short-term laggard between two 3x funds or the relative strength of 2x Nasdaq-100 and 2x S&P 500 funds. There is no stock universe to sort, so they cannot be validated cross-sectionally; their evidence is the time-series backtest and the economic logic of the regime rule. They tend to be more concentrated (often a single holding) and more dispersed in their results, as the scatter at the top of this page shows: the ETF-timing points span CAGRs from 24.3% to 51.9%, while the stock-factor group sits between 27.2% and 32.0%.

Risk-adjusted return: how the factors compare

Every strategy here uses leverage, so raw CAGR rewards whoever took the most risk in a bull market. The Sharpe ratio is the cleaner lens; the figures below are full-window monthly Sharpe ratios with finlab's 2% risk-free deduction, as in the table above, and QQQ is measured the same way.

All ten clear QQQ's 0.98 on this basis, but not by the same distance. Breakout/trend (1.04) and the aggressive ETF rotation (1.12) clear it only narrowly; without the risk-free deduction the picture is the same, with breakout at 1.11 against QQQ's 1.09. The stock-factor group lands in a 1.27 to 1.47 band, with earnings surprise (1.47), free cash flow (1.42), and value + quality (1.37) on top. That ranking cannot be credited to fundamentals: momentum, which uses no fundamental data, scores 1.36, level with value + quality, and all five hold about half the book in a defensive ETF during risk-on months, which damps their volatility. How much of the higher Sharpe comes from the factor and how much from that blend cannot be separated in these backtests. The ETF-timing group sits lower, from 1.04 to 1.24. Higher headline CAGR did not buy a higher Sharpe: short-term reversal and the aggressive ETF rotation earned the two largest returns (51.9% and 41.0%) but ran the highest volatility in the family (34.1% and 36.2% a year) and two of the three deepest drawdowns, so their Sharpe ratios are middling (1.24) to low (1.12).

Drawdown: the shared stop shows up in the data

Two things stand out. First, all ten cut their worst loss below QQQ's -35.1% buy-and-hold drawdown, even while holding 2x and 3x products: the regime gate is doing real work in weak markets. Second, the five stock-factor strategies share an identical -21.1% floor. That is the same 8% stop on the same leveraged ETFs, held in the same half-defensive blend, producing the same worst case: a property of the engine rather than something to credit to any one factor. The ETF-timing strategies that switch on faster signals ran deeper drawdowns: -27.8% for short-term reversal, -27.9% for breakout/trend, and -32.4% for the aggressive ETF rotation, which holds a single 3x fund behind the widest stop (10%) and came closest to QQQ. The conservative ETF rotation, holding two 2x funds behind a 6% stop, had the shallowest drawdown of all at -19.3% and the lowest CAGR (24.3%), a direct illustration of the trade-off.

Does it survive out of sample?

Every strategy was calibrated on 2016-2021 and then left untouched on 2022-present, an unseen stretch that includes the 2022 bear market.

The refreshed data is less flattering here than the June version was. Nine of the ten earned a lower CAGR after 2021 than in the research period; only quality growth rose (26.4% to 28.1%). The ETF-timing group took the largest hits: short-term reversal fell from 63.3% to 39.6%, the aggressive ETF rotation from 49.1% to 31.2%, and the conservative ETF rotation from 32.6% to 14.8%, with its out-of-sample monthly Sharpe down to 0.95. Two facts keep that in proportion. QQQ itself compounded at only 13.3% a year over 2022-present with a 0.82 monthly Sharpe (no risk-free deduction, the same basis as the strategies' out-of-sample Sharpe), and all ten strategies stayed above both figures. And the stock-factor group held its risk-adjusted return: monthly Sharpe of 1.50 to 1.53 out of sample, against 0.95 to 1.16 for the ETF-timing five.

There is an important caveat the chart makes obvious: momentum, quality growth, value + quality, and free cash flow post the same out-of-sample CAGR (28.1%) and Sharpe (1.50), with earnings surprise close behind (29.2%, 1.53). That is the shared-execution design surfacing again. In a sustained risk-on regime, different breadth factors agree that conditions are supportive, so they all hold the same leveraged ETFs and their out-of-sample returns converge. Read positively, the edge does not depend on which factor you pick; read skeptically, these strategies are less diversifying against each other than their names suggest, and running several of them is not genuine diversification.

Robustness checks: neighbouring parameters, two halves, trading costs

Each strategy study now carries three robustness checks. "Beats QQQ" means a higher daily Sharpe ratio (mean daily return divided by its standard deviation, annualized with √252, no risk-free rate) than buying and holding QQQ with dividends over exactly the same dates. The neighbouring-parameter check moves the two main parameters (usually a signal threshold or lookback, plus the length of the QQQ trend average) one step either side of the original and needs at least 7 of the 9 combinations to beat QQQ. The two-halves check cuts the window by trading days at 2021-07-08 and needs each half to beat QQQ on its own (QQQ scored 1.20 in the first half and 0.71 in the second). The cost check charges 0.1% on every buy and every sell for spread and slippage, since finlab's US defaults charge no commission or tax. A strategy passes only if it passes all three.

Seven of the eight strategies checked pass all three. Every one of them stays ahead of QQQ after paying 0.1% per side, with CAGR falling by about 1.0 to 2.2 points, and every one keeps at least seven of nine neighbouring settings ahead of QQQ, although the aggressive ETF rotation does so with no room to spare: its two misses both use the shortest, three-month momentum lookback. Breakout/trend fails the two-halves check. Over the first half, 2016-05-02 to 2021-07-08, its daily Sharpe was 1.16 against QQQ's 1.20, so its full-window lead comes from the second half (1.11 against 0.71), when QQQ itself was weak. For a reader, that means the breakout system has not shown it can beat simply holding QQQ through a strong bull market; it is reported as a failure, and its parameters were not changed to make it pass. Quality growth and value + quality were not checked on their own, because the quality-value study ran the checks on its free-cash-flow version. Each study shows the full grid, the two halves, and the cost run, and python strategy.py --validate with that study's downloadable script reproduces them.

What changed since the June version

The June version of this page covered June 2016 to June 2026. Every number has since been re-run on FinLab data through 2026-10-26 with finlab 2.2.20, and the history moved, not only the last four months:

• Price corrections. FinLab corrected its US prices in October 2026: missing dividends were added to adjusted prices on 2026-10-10, and split and spin-off adjustments were fixed. finlab 2.2.20 also changed how a stop-loss that triggers on a rebalance day is costed.
• Liquid-universe fix (strategies 01, 03, 04, 05, 08). FinLab's US volume is not split-adjusted, while the adjusted close is. After the split corrections, adjusted close times volume overstated a stock's dollar volume before a reverse split by the split ratio, so tiny reverse-split names were entering the top-500 universe and passing the $10 price floor. The five stock-factor strategies now use the unadjusted close for dollar volume and the price floor, and still use the adjusted close for returns.

The net effect: every CAGR is lower than in June, by 2.8 points (low volatility) to 15.5 points (short-term reversal), and every full-window monthly Sharpe is lower too. When the family was assembled, each strategy had to clear a CAGR above 30% and a monthly Sharpe above 1.1 over the full window. On the refreshed data five no longer do: momentum (29.9%), quality growth (27.2%), low volatility (29.3%), breakout/trend (29.8%, with a 1.04 Sharpe), and the conservative ETF rotation (24.3%). They are reported as they are; no parameters were changed to bring them back over the line.

Copy this prompt to your AI agent

FinLab's setup path is now one prompt. Paste it into Codex app or Claude cowork, and the AI will read https://finlab.finance/en/setup, install the FinLab skill when needed, then run or modify the strategy in this article.

Help me set up FinLab and compare US quant trading strategies: https://finlab.finance/en/setup

The setup page is https://finlab.finance/en/setup.

The leveraged-ETF risk you must understand

Every strategy in this family holds leveraged ETFs, so they share the same core risks:

• Daily reset and path dependency. A 3x daily ETF compounds daily, so a choppy, sideways market erodes value even if the index ends flat. This is why none of these strategies hold leverage unconditionally.
• Gaps can beat stops. The 6% to 10% stop is a touched level in the backtest; a large overnight gap can fill worse in live trading.
• Monthly switching is not crash protection. A fast crash inside a month is taken at full leverage until the stop or the next rebalance.
• Concentration. Holding one to three ETFs means single-position and single-theme (large-cap US growth) risk. The higher-return ETF-timing strategies are the most concentrated.

If those risks are unacceptable to you, the individual studies also report each factor at the stock level, without leverage. In the momentum study, for example, the top momentum decile beat the bottom decile by 22.1% a year on average as a long-short spread before costs. None of those decile portfolios is the strategy tested on this page, and none of them carries the drawdown figures above.

Backtest method and limits

The assumptions are shared across the family:

The publication bar for these studies is a risk-adjusted ratio of at least 1.5. On finlab's full-window monthly Sharpe (2% risk-free rate) no strategy reaches it any more; the highest is earnings surprise at 1.47, or 1.57 without the risk-free deduction, where free cash flow (1.52) also clears it. Over 2022-present (no deduction) the five stock-factor strategies sit at 1.50 to 1.53, and the five ETF-timing strategies, at 0.95 to 1.16, do not. The lower figures are reported as they are rather than reframed.

Where the edge comes from

Each signal rests on a documented anomaly rather than a fitted curve. Cross-sectional momentum was established by Jegadeesh & Titman (1993) and built into a standard factor by Carhart (1997). The earnings-surprise strategy draws on post-earnings-announcement drift, documented by Bernard & Thomas (1989). The quality and free-cash-flow strategies rest on the gross-profitability premium of Novy-Marx (2013), and the low-volatility strategy on the low-beta effect formalized by Frazzini & Pedersen (2014). That momentum and value persist together across markets is the broader evidence in Asness, Moskowitz & Pedersen (2013). The contribution here is not the anomalies; it is expressing them through a disciplined leveraged-ETF regime gate and measuring every version on the same data, window, and benchmark.

Read the individual strategy write-ups

Each study has full validation, the three robustness checks, downloadable code, and an interactive report:

• US momentum strategy: the first full study, with decile sorts, rank IC, breadth as a regime gate, and the leveraged-ETF execution in detail. Momentum performance and Studio.
• Post-earnings announcement drift strategy: a 19,933-event earnings surprise study, why single-name PEAD barely survives in liquid large caps, and how surprise breadth still works as a regime signal. Earnings surprise performance and Studio.
• Quality and value strategy: three fundamental screens (quality growth, value plus quality, free cash flow) run through decile and rank-IC tests. They are weak as stock pickers, strong as breadth gates, and all three converge to one trade out of sample. Quality-value performance and Studio.
• ETF rotation strategy: one QQQ risk-on/risk-off signal at two leverage budgets (conservative 2x, aggressive 3x), why the signal filters volatility rather than predicting returns, and what that buys for surviving leveraged-ETF compounding. ETF rotation performance and Studio.
• Short-term mean reversion strategy: the family's highest CAGR (51.9%), from a 3-day laggard rule between TQQQ and TECL inside the regime gate, with an event study showing the reversal tilt is real but statistically weak (t-stat 1.25), so the regime engine does most of the work. Mean reversion performance and Studio.
• Breakout trend following strategy: fresh QQQ 21-day highs gated by a 200-day trend filter. In its event study, filtered highs followed through 80% of the time at three months, yet it has the lowest monthly Sharpe in the family (1.04, just above QQQ's 0.98) and is the one strategy that fails the two-halves check.
• Low volatility strategy: a QQQ regime filter expressed through one 2x leveraged ETF, and the most candid result in the family. Naive QLD out-earned it on raw CAGR (33.3% vs 29.3%), so what the regime filter actually bought was the risk column, cutting the drawdown to -22% from -64% by forecasting volatility rather than crashes. Low volatility performance and Studio.

For definitions of the metrics used above, see the glossary. To see what teams build with finlab, browse the use cases, the research blog, and the team behind FinLab.

Tested approaches that did not hold up

• TQQQ's 200-day moving average and leveraged ETF timing: none of seven rules cleared all three checks; TQQQ with the 200-day filter returned 45.1% a year with a -60.9% drawdown, but only 4 of 9 neighbouring settings beat QQQ and it lost the first half.
• The Santa Claus rally and calendar effects: none of five calendar effects or their 90 neighbouring settings beat QQQ, and the Santa Claus rally's seven-day window gained just 0.15% on average.
• The Rule of 40, Magic Formula and other stock screens: none of six screens counts as a working strategy in the study; five lost across all 63 neighbouring settings, while dividend growth trailed QQQ from four of seven start dates and SPY from all seven.
• 60/40, All Weather and other lazy portfolios: all seven had shallower drawdowns than QQQ's -35.1%, with Faber's at -9.0%, but none beat QQQ on risk-adjusted return across all three robustness checks.

Who this is for

This comparison fits an investor who wants a rules-based, growth-tilted US approach, understands leveraged ETFs, and can tolerate 20%+ drawdowns in exchange for higher compounding. It does not fit anyone who needs capital stability, cannot monitor a monthly rebalance, or is uncomfortable holding 2x or 3x products at any time.

FAQ

What are these US quant trading strategies, in one sentence? Ten rules-based systems that hold leveraged growth ETFs (TQQQ, TECL, QLD, SSO) only while a shared QQQ trend gate is open, six of them with an extra breakout or breadth condition, and rotate to a defensive ETF otherwise, rebalanced monthly with a 6% to 10% stop-loss.

Why do so many of them have similar drawdowns? Because they share the same execution: the five stock-factor strategies hold the same leveraged ETFs behind the same 8% stop, usually as a quarter each next to a defensive ETF at about half the book. The extra conditions change the timing at the margin, not the instrument, so the worst-case drawdown is largely a property of the engine.

Do all ten beat QQQ? Over the full window, yes: every one compounded faster than QQQ's 20.7%, fell less than its -35.1%, and had a higher monthly Sharpe measured the same way, although breakout/trend's lead is thin (1.04 against 0.98). The robustness checks are stricter because they also compare sub-periods: seven of the eight strategies checked pass, and breakout/trend fails because it trailed QQQ in the first half of the window.

Which one is "best"? There is no single best. Earnings surprise had the highest full-window Sharpe (1.47) and passed all three robustness checks, short-term reversal the highest CAGR (51.9%) at a -27.8% drawdown, and the conservative ETF rotation the shallowest drawdown (-19.3%) at the lowest CAGR (24.3%). The right choice depends on how much drawdown you can hold.

Are the returns realistic given no transaction costs? The default run charges no commission, tax, or slippage. With 0.1% charged on every buy and every sell, every strategy checked still beat QQQ's daily Sharpe, and CAGR fell by about 1.0 to 2.2 points. Leveraged-ETF spreads and overnight gaps through the stop can still make live results worse than either figure.

Can I run and modify these myself? Yes. Use the AI-assisted setup flow and ask your agent to build any version, or download a strategy script from its study and run it after setup.

Do I need single-stock data or just ETF data? The stock-factor strategies need us_price:* for the screen (adjusted close for returns, unadjusted close and volume for the liquidity filter) and us_fund_price:* for the ETFs; the ETF-timing strategies need only us_fund_price:*. Both are in a standard FinLab account.

Last updated: 2026-10 | Backtest window: 2016-05 to 2026-10 | Benchmark: QQQ total return | Author: FinLab Quant Research (reviewed by a quantitative researcher)

Investing involves risk, and past performance does not represent future results. Leveraged ETFs carry additional risks including volatility decay and amplified losses. This content is for educational purposes only and is not investment advice and does not constitute investment advice; evaluate any strategy against your own risk tolerance.